In a MVVM application with EF Core as ORM I decided to model a table with a manually inserted, textual primary key.
This is because in this specific application I'd rather use meaningful keys instead of meaningless integer ids, at least for simple key-value tables like the table of the countries of the world.
I have something like:
Id | Description
-----|--------------------------
USA | United States of America
ITA | Italy
etc. etc.
So the entity is:
public class Country
{
[Key]
[DatabaseGenerated(DatabaseGeneratedOption.None)]
public string Id { get; set; }
public string Description { get; set; }
}
Here is my viewmodel. It's little more than a container for the ObservableCollection of Countries. Actually it gets loaded from a repository. It's trivial and I inlcuded the entire code at the end. It's not really relevant and I could do with just the DbContext as well. But I wanted to show all the layers to see where the solution belongs to. Oh yes, then it contains the synchronizing code that actually offends EF Core.
public class CountriesViewModel
{
//CountryRepository normally would be injected
public CountryRepository CountryRepository { get; set; } = new CountryRepository(new AppDbContext());
public ObservableCollection<Country> Countries {get; set;}
public CountriesViewModel()
{
Countries = new ObservableCollection<Country>();
Countries.CollectionChanged += Countries_CollectionChanged;
}
private void Countries_CollectionChanged(object sender, System.Collections.Specialized.NotifyCollectionChangedEventArgs e)
{
foreach (Country c in e.NewItems)
{
CountryRepository.Add(c);
}
}
}
In my MainWindow I just have:
<Window.DataContext>
<local:CountriesViewModel/>
</Window.DataContext>
<DockPanel>
<DataGrid ItemsSource="{Binding Countries}"/>
</DockPanel>
Problem and question
Now this doesn't work. When we try to insert a new record, in this case I do it using the automatic feature of DataGrid I get a:
System.InvalidOperationException: 'Unable to track an entity of type 'Country'
because primary key property 'Id' is null.'
Each time i add a new record to the ObservableCollection I also try to add it back to the repository, that in turn adds it on the EF DbContext that doesn't accept entities with null key.
So what are my options here?
One is postponing the addition of the new record till the Id has been inserted. This is not trivial as the collection handling that I've shown, but this is not the problem. The worst is that this way I would have some record that are tracked by EF (the updated and the deleted and the new with pk assigned) and some that are tracked by the view model (the new ones with the key not yet assigned).
Another is using alternate keys; I would have an integer, autogenerated primary key and the ITA,USA etc code would be an alternate key that would be used also in relations. It's not so bad from as simplicity, but I'd like a application-only solution.
What I'm looking for
I'm looking for a neat solution here, a pattern to be used whenever this problem arises and that plays well in the context of a MVVM/EF application.
Of course I could also look in the direction of the view events, that is force the user to insert the key before of a certain event that triggers the insertion. I would consider it a second-class solution because it is sort of view dependent.
Remaining code
Just for completeness, in case that you want to run the code, here is the remaining code.
DbContext
(Configured for postgres)
public class AppDbContext : DbContext
{
protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)
{
optionsBuilder.UseNpgsql("Host=localhost;Database=WpfApp1;Username=postgres;Password=postgres");
}
public DbSet<Country> Countries { get;set; }
}
Repository
The reason why I implemented the repository for such a simple example is because I think that a possible solution may be to include the new-without-key records managment in the Repository instead of in the viewmodel. I still hope that someone comes out with a simpler solution.
public class CountryRepository
{
private AppDbContext AppDbContext { get; set; }
public CountryRepository(AppDbContext appDbContext) => AppDbContext = appDbContext;
public IEnumerable<Country> All() => AppDbContext.Countries.ToList();
public void Add(Country country) => AppDbContext.Add(country);
//ususally we don't have a save here, it's in a Unit of Work;
//for the example purpose it's ok
public int Save() => AppDbContext.SaveChanges();
}
Probably the cleanest way to address the aforementioned issue in EF Core is to utilize temporary value generation on add. In order to do that, you would need a custom ValueGenerator like this:
using Microsoft.EntityFrameworkCore.ChangeTracking;
using Microsoft.EntityFrameworkCore.ValueGeneration;
public class TemporaryStringValueGenerator : ValueGenerator<string>
{
public override bool GeneratesTemporaryValues => true; // <-- essential
public override string Next(EntityEntry entry) => Guid.NewGuid().ToString();
}
and fluent configuration similar to this:
modelBuilder.Entity<Country>().Property(e => e.Id)
.HasValueGenerator<TemporaryStringValueGenerator>()
.ValueGeneratedOnAdd();
The potential drawbacks are:
In pre EF Core 3.0 the generated temporary value is set onto entity instance, thus would be visible in the UI. This has been fixed in EF Core 3.0, so now Temporary key values are no longer set onto entity instances
Even though the property looks empty (null) and is required (default for primary/alternate keys), if you don't provide explicit value, EF Core will try to issue INSERT command and read the "actual" value back from database similar to identity and other database generated values, which in this case will lead to non user friendly database generated runtime exception. But EF Core in general does not do validations, so this won't be so different - you have to add and validate property required rule in the corresponding layer.
This could be a duplicate question but a lot of searching for the words in the title only got me a lot of unrelated results.
I have an entity that's roughly set up like this:
public abstract class A
{
public GUID AId { get; set; }
public string SomeProperty { get; set; }
}
public class B : A
{
public string SomeOtherProperty { get; set; }
}
The context has public DbSet<B> BInstances { get; set; } for B objects. In OnModelCreating, the mapping has A set to ignored and B is mapped to a table called TableB.
The AId field is not auto-generated (not an identity field) but it's set to be primary key, both in the database and in the mapping. In the database, the field is defined as a non-null uniqueidentifier with no default.
At runtime, I'm loading an instance of B using its key (_token is just a CancellationToken):
var b = await (dbCtx.BInstances.FirstOrDefaultAsync(e => e.AId), _token));
Then, a property of b is set and I try to save it back to database:
b.SomeOtherProperty = "some new text";
await (dbCtx.SaveChangesAsync(_token));
At this point, I'm getting a Violation of PRIMARY KEY constraint error from the database, stating that the value of AId cannot be inserted because it'd be a duplicate. Of course, the ID is already in the database, I loaded the entity from there, using the ID. For some reason, EF generates an INSERT statement, not an UPDATE and I don't understand why.
When I check dbCtx.Entry(b).State, it's already set to EntityState.Modified. I'm at a loss - can someone point out what I'm doing wrong? I never had issues with updating entities before but I haven't used EF with GUID primary keys (usually I use long primary keys).
I'm using EF 6 and .NET Framework 4.7.1.
Thank you all for the suggestions - this turned out to be a mapping problem that I caused.
In my OnModelCreating() call, I called MapInheritedProperties() on a type that didn't inherit from a base class (other than object, of course) - this seems to have triggered a problem. Other entities that do share a base class worked fine with the mapping call.
I also called ToTable() directly against the entity class - this broke my table mapping for reasons I do not understand. Once I moved that call inside Map(), it started working as expected.
So I went from this:
entity.ToTable("tablename");
to this:
entity.Map(m => m.ToTable("tablename"));
to solve the problem.
Hopefully this will be useful for future readers.
try this
b.SomeOtherProperty = "some new text";
dbCtx.BInstances.AddOrUpdate(b);
await (dbCtx.SaveChangesAsync(_token));
AddorUpdate will update your b instance if it is already added.
I have a mental debate with myself every time I start working on a new project and I am designing my POCOs. I have seen many tutorials/code samples that seem to favor foreign key associations:
Foreign key association
public class Order
{
public int ID { get; set; }
public int CustomerID { get; set; } // <-- Customer ID
...
}
As opposed to independent associations:
Independent association
public class Order
{
public int ID { get; set; }
public Customer Customer { get; set; } // <-- Customer object
...
}
I have worked with NHibernate in the past, and used independent associations, which not only feel more OO, but also (with lazy loading) have the advantage of giving me access to the whole Customer object, instead of just its ID. This allows me to, for example, retrieve an Order instance and then do Order.Customer.FirstName without having to do a join explicitly, which is extremely convenient.
So to recap, my questions are:
Are there any significant disadvantages in
using independent associations? and...
If there aren't any, what
would be the reason of using foreign key associations at all?
If you want to take full advantage of ORM you will definitely use Entity reference:
public class Order
{
public int ID { get; set; }
public Customer Customer { get; set; } // <-- Customer object
...
}
Once you generate an entity model from a database with FKs it will always generate entity references. If you don't want to use them you must manually modify the EDMX file and add properties representing FKs. At least this was the case in Entity Framework v1 where only Independent associations were allowed.
Entity framework v4 offers a new type of association called Foreign key association. The most obvious difference between the independent and the foreign key association is in Order class:
public class Order
{
public int ID { get; set; }
public int CustomerId { get; set; } // <-- Customer ID
public Customer Customer { get; set; } // <-- Customer object
...
}
As you can see you have both FK property and entity reference. There are more differences between two types of associations:
Independent association
It is represented as separate object in ObjectStateManager. It has its own EntityState!
When building association you always need entitites from both ends of association
This association is mapped in the same way as entity.
Foreign key association
It is not represented as separate object in ObjectStateManager. Due to that you must follow some special rules.
When building association you don't need both ends of association. It is enough to have child entity and PK of parent entity but PK value must be unique. So when using foreign keys association you must also assign temporary unique IDs to newly generated entities used in relations.
This association is not mapped but instead it defines referential constraints.
If you want to use foreign key association you must tick Include foreign key columns in the model in Entity Data Model Wizard.
Edit:
I found that the difference between these two types of associations is not very well known so I wrote a short article covering this with more details and my own opinion about this.
Use both. And make your entity references virtual to allow for lazy loading. Like this:
public class Order
{
public int ID { get; set; }
public int CustomerID { get; set; }
public virtual Customer Customer { get; set; } // <-- Customer object
...
}
This saves on unnecessary DB lookups, allows lazy loading, and allows you to easily see/set the ID if you know what you want it to be. Note that having both does not change your table structure in any way.
Independent association doesn't work well with AddOrUpdate that is usually used in Seed method. When the reference is an existing item, it will be re-inserted.
// Existing customer.
var customer = new Customer { Id = 1, Name = "edit name" };
db.Set<Customer>().AddOrUpdate(customer);
// New order.
var order = new Order { Id = 1, Customer = customer };
db.Set<Order>().AddOrUpdate(order);
The result is existing customer will be re-inserted and new (re-inserted) customer will be associated with new order.
Unless we use the foreign key association and assign the id.
// Existing customer.
var customer = new Customer { Id = 1, Name = "edit name" };
db.Set<Customer>().AddOrUpdate(customer);
// New order.
var order = new Order { Id = 1, CustomerId = customer.Id };
db.Set<Order>().AddOrUpdate(order);
We have the expected behavior, existing customer will be associated with new order.
I favour the object approach to avoid unnecessary lookups. The property objects can be just as easily populated when you call your factory method to build the whole entity (using simple callback code for nested entities). There are no disadvantages that I can see except for memory usage (but you would cache your objects right?). So, all you are doing is substituting the stack for the heap and making a performance gain from not performing lookups. I hope this makes sense.
Background
I have a class that looks more or less like this:
public class MyClass
{
[Id]
public long Id { get; set; }
public string MyProperty { get; set; }
public bool MyBoolean { get; set; }
public string AnotherProperty { get; set; }
public MyClass ChildOne { get; set; }
public MyClass ChildTwo { get; set; }
}
I will need to use a stored procedure to load a set of records, but that's ok as long as the structure itself is correct.
For any instance of MyClass, one or both of the children can be null. Any instance of MyClass can be used in a parent class - but the child itself doesn't need to know about this relationship, and a child can be used by any number of parents.
Problem
With this structure, I get the following error when creating a new migration:
Unable to determine the principal end of an association between the
types 'MyClass' and 'MyClass'. The principal end of this association
must be explicitly configured using either the relationship fluent API
or data annotations.
This error makes sense - when given a structure of an object with a foreign key to itself, I am not surprised that EF has a hard time determining the principal end. I'm not sure how to fix this, though.
I've tried some different Fluent mappings:
modelBuilder.Entity<MyClass>().HasOptional(x => x.ChildOne).WithOptionalPrincipal(x => x.ChildOne);
modelBuilder.Entity<MyClass>().HasOptional(x => x.ChildOne).WithOptionalDependent(x => x.ChildOne);
modelBuilder.Entity<MyClass>().HasOptional(x => x.ChildOne);
(Note: I didn't try these concurrently - I did one at a time & duplicated it for ChildTwo.)
I was able to get a migration to work by adding a ChildThree property to MyClass, but that doesn't make sense and isn't a useful property; it just creates another foreign key on the table but this isn't needed in my model.
So, in summary:
How do I get this structure to work the way I want? I think the secret is in some Fluent mapping voodoo but I'm very unfamiliar with that library and I don't know how to get that to work.
Why does adding a third (unneeded, unwanted) property fix everything and allow the migration to scaffold?
Your fluent mapping is totally wrong.
You should do something like this:
modelBuilder.Entity<MyClass>().HasOptional(p => p.ChildOne).WithOptionalDependent();
modelBuilder.Entity<MyClass>().HasOptional(p => p.ChildTwo).WithOptionalDependent();
I'm having problems setting up an Entity Framework 4 model.
A Contact object is exposed in the database as an updateable view. Also due to the history of the database, this Contact view has two different keys, one from a legacy system. So some other tables reference a contact with a 'ContactID' while other older tables reference it with a 'LegacyContactID'.
Since this is a view, there are no foreign keys in the database, and I'm trying to manually add associations in the designer. But the fluent associations don't seem to provide a way of specifying which field is referenced.
How do I build this model?
public class vwContact
{
public int KeyField { get; set; }
public string LegacyKeyField { get; set; }
}
public class SomeObject
{
public virtual vwContact Contact { get; set; }
public int ContactId { get; set; } //references vwContact.KeyField
}
public class LegacyObject
{
public virtual vwContact Contact { get; set; }
public string ContactId { get; set; } //references vwContact.LegacyKeyField
}
ModelCreatingFunction(modelBuilder)
{
// can't set both of these, right?
modelBuilder.Entity<vwContact>().HasKey(x => x.KeyField);
modelBuilder.Entity<vwContact>().HasKey(x => x.LegacyKeyField);
modelBuilder.Entity<LegacyObject>().HasRequired(x => x.Contact).???
//is there some way to say which key field this reference is referencing?
}
EDIT 2: "New things have come to light, man" - His Dudeness
After a but more experimentation and news, I found using a base class and child classes with different keys will not work by itself. With code first especially, base entities must define a key if they are not explicitly mapped to tables.
I left the suggested code below because I still recommend using the base class for your C# manageability, but I below the code I have updated my answer and provided other workaround options.
Unfortunately, the truth revealed is that you cannot accomplish what you seek without altering SQL due to limitations on EF 4.1+ code first.
Base Contact Class
public abstract class BaseContact
{
// Include all properties here except for the keys
// public string Name { get; set; }
}
Entity Classes
Set this up via the fluent API if you like, but for easy illustration I've used the data annotations
public class Contact : BaseContact
{
[Key]
public int KeyField { get; set; }
public string LegacyKeyField { get; set; }
}
public class LegacyContact : BaseContact
{
public int KeyField { get; set; }
[Key]
public string LegacyKeyField { get; set; }
}
Using the Entities
Classes that reference or manipulate the contact objects should reference the base class much like an interface:
public class SomeCustomObject
{
public BaseContact Contact { get; set; }
}
If later you need to programmatically determine what type you are working with use typeof() and manipulate the entity accordingly.
var co = new SomeCustomObject(); // assume its loaded with data
if(co.Contact == typeof(LegacyContact)
// manipulate accordingly.
New Options & Workarounds
As I suggested in comment before, you won't be able to map them to a single view/table anyway so you have a couple options:
a. map your objects to their underlying tables and alter your "get/read" methods on repositories and service classes pull from the joined view -or-
b. create a second view and map each object to their appropriate view.
c. map one entity to its underlying table and one to the view.
Summary
Try (B) first, creating a separate view because it requires the least amount of change to both code and DB schema (you aren't fiddling with underlying tables, or affecting stored procedures). It also ensures your EF C# POCOs will function equivalently (one to a view and one to table may cause quirks). Miguel's answer below seems to be roughly the same suggestion so I would start here if it's possible.
Option (C) seems worst because your POCO entities may behave have unforseen quirks when mapped to different SQL pieces (tables vs. views) causing coding issues down the road.
Option (A), while it fits EF's intention best (entities mapped to tables), it means to get your joined view you must alter your C# services/repositories to work with the EF entities for Add, Update, Delete operations, but tell the Pull/Read-like methods to grab data from the joint views. This is probably your best choice, but involves more work than (B) and may also affect Schema in the long run. More complexity equals more risk.
Edit I'm not sure this is actually possible, and this is why:
The assumption is that a foreign key references a primary key. What you've got is two fields which are both acting as primary keys of vwContact, but depending on which object you ask it's a different field that's the primary key. You can only have one primary key at once, and although you can have a compound primary key you can't do primary key things with only half of it - you have to have a compound foreign key with which to reference it.
This is why Entity Framework doesn't have a way to specify the mapping column on the target side, because it has to use the primary key.
Now, you can layer some more objects on top of the EF entities to do some manual lookup and simulate the navigation properties, but I don't think you can actually get EF to do what you want because SQL itself won't do what you want - the rule is one primary key per table, and it's not negotiable.
From what you said about your database structure, it may be possible for you to write a migration script which can give the contact entities a consistent primary key and update everything else to refer to them with that single primary key rather than the two systems resulting from the legacy data, as you can of course do joins on any fields you like. I don't think you're going to get a seamlessly functional EF model without changing your database though.
Original Answer That Won't Work
So, vwContact contains a key KeyField which is referenced by many SomeObjects and another key LegacyKeyField which is referenced by many LegacyObjects.
I think this is how you have to approach this:
Give vwContact navigation properties for SomeObject and LegacyObject collections:
public virtual ICollection<SomeObject> SomeObjects { get; set; }
public virtual ICollection<LegacyObject> LegacyObjects { get; set; }
Give those navigation properties foreign keys to use:
modelBuilder.Entity<vwContact>()
.HasMany(c => c.SomeObjects)
.WithRequired(s => s.Contact)
.HasForeignKey(c => c.KeyField);
modelBuilder.Entity<vwContact>()
.HasMany(c => c.LegacyObjects)
.WithRequired(l => l.Contact)
.HasForeignKey(c => c.LegacyKeyField);
The trouble is I would guess you've already tried this and it didn't work, in which case I can't offer you much else as I've not done a huge amount of this kind of thing (our database is much closer to the kinds of thing EF expects so we've had to do relatively minimal mapping overrides, usually with many-to-many relationships).
As for your two calls to HasKey on vwContact, they can't both be the definitive key for the object, so it's either a compound key which features both of them, or pick one, or there's another field you haven't mentioned which is the real primary key. From here it's not really possible to say what the right option there is.
You should be able to do this with two different objects to represent the Contact view.
public class vwContact
{
public int KeyField { get; set; }
public string LegacyKeyField { get; set; }
}
public class vwLegacyContact
{
public int KeyField { get; set; }
public string LegacyKeyField { get; set; }
}
public class SomeObject
{
public virtual vwContact Contact { get; set; }
public int ContactId { get; set; } //references vwContact.KeyField
}
public class LegacyObject
{
public virtual vwLegacyContact Contact { get; set; }
public string ContactId { get; set; } //references vwLegacyContact.LegacyKeyField
}
ModelCreatingFunction(modelBuilder)
{
// can't set both of these, right?
modelBuilder.Entity<vwContact>().HasKey(x => x.KeyField);
modelBuilder.Entity<vwLegacyContact>().HasKey(x => x.LegacyKeyField);
// The rest of your configuration
}
I have tried everything that you can imagine, and found that most solutions won't work in this version of EF... maybe in future versions it supports referencing another entity by using an unique field, but this is not the case now. I also found two solutions that work, but they are more of a workaround than solutions.
I tried all of the following things, that didn't work:
Mapping two entities to the same table: this is not allowed in EF4.
Inheriting from a base that has no key definitions: all root classes must have keys, so that inherited classes share this common key... that is how inheritance works in EF4.
Inheriting from base class that defines all fields, including keys, and then use modelBuilder to tell wich base-properties are keys of the derived types: this doesn't work, because the methos HasKey, Property and others that take members as parameters, must reference members of the class itself... referencing properties of a base class is not allowed. This cannot be done: modelBuilder.HasKey<MyClass>(x => x.BaseKeyField)
The two things that I did that worked:
Without DB changes: Map to the table that is source of the view in question... that is, if vwContact is a view to Contacts table, then you can map a class to Contacts, and use it by setting the key to the KeyField, and another class mapping to the vwContacts view, with the key being LegacyKeyField. In the class Contacts, the LegacyKeyField must exist, and you will have to manage this manually, when using the Contacts class. Also, when using the class vwContacts you will have to manually manage the KeyField, unless it is an autoincrement field in the DB, in this case, you must remove the property from vwContacts class.
Changing DB: Create another view, just like the vwContacts, say vwContactsLegacy, and map it to a class in wich the key is the LegacyKeyField, and map vwContacts to the original view, using KeyField as the key. All limitations from the first case also applies: the vwContacts must have the LegacyKeyField, managed manually. And the vwContactsLegacy, must have the KetField if it is not autoincrement idenitity, otherwise it must not be defined.
There are some limitations:
As I said, these solutions are work-arounds... not real solutions, there are some serious implications, that may even make them undesirable:
EF does not know that you are mapping two classes to the same thing. So when you update one thing, the other one could be changed or not, it depends if the objects is cached or not. Also, you could have two objects at the same time, that represents the same thing on the backing storage, so say you load a vwContact and also a vwContactLegacy, changes both, and then try to save both... you will have to care about this yourself.
You will have to manage one of the keys manually. If you are using vwContacts class, the KeyFieldLegacy is there, and you must fill it. If you want to create a vwContacts, and associate is with a LegacyObject, then you need to create the reference manually, because LegacyObject takes a vwContactsLegacy, not a vwContacts... you will have to create the reference by setting the ContactId field.
I hope that this is more of a help than a disillusion, EF is a powerfull toy, but it is far from perfect... though I think it's going to get much better in the next versions.
I think this may be possible using extension methods, although not directly through EF as #Matthew Walton mentioned in his edit above.
However, with extension methods, you can specify what to do behind the scenes, and have a simple call to it.
public class LegacyObject
{
public virtual vwContact Contact { get; set; }
public string ContactId { get; set; } //references vwContact.LegacyKeyField
}
public class LegacyObjectExtensions
{
public static vwContact Contacts(this LegacyObject legacyObject)
{
var dbContext = new LegacyDbContext();
var contacts = from o in legacyObject
join c in dbContext.vwContact
on o.ContactId == c.LegacyKeyField
select c;
return contacts;
}
}
and
public class SomeObject
{
public virtual vwContact Contact { get; set; }
public int ContactId { get; set; } //references vwContact.KeyField
}
public class SomeObjectExtensions
{
public static vwContact Contacts(this SomeObject someObject)
{
var dbContext = new LegacyDbContext();
var contacts = from o in someObject
join c in dbContext.vwContact
on o.ContactId == c.KeyField
select c;
return contacts;
}
}
Then to use you can simply do like this:
var legacyContacts = legacyObject.Contacts();
var someContacts = someObject.Contacts();
Sometimes it makes more sense to map it from the other end of the relationship, in your case:
modelBuilder.Entity<LegacyObject>().HasRequired(x => x.Contact).WithMany().HasForeignKey(u => u.LegacyKeyField);
however this will require that u.LegacyKeyField is marked as a primary key.
And then I'll give my two cents:
if the Legacy db is using LegacyKeyField, then perhaps the legacy db will be read only. In this case we can create two separate contexts Legacy and Non-legacy and map them accordingly. This can potentially become a bit messy as you'd have to remember which object comes from which context. But then again, nothing stops you from adding the same EF code first object into 2 different contexts
Another solution is to use views with ContactId added for all other legacy tables and map them into one context. This will tax performance for the sake of having cleaner context objects, but this can be counteracted on sql side: indexed views, materialized views, stored procs, etc. So than LEGACY_OBJECT becomes VW_LEGACY OBJECT with CONTACT.ContactId brought over, then:
modelBuilder.Entity<LegacyObject>().ToTable("VW_LEGACY_OBJECT");
modelBuilder.Entity<LegacyObject>().HasRequired(x => x.Contact).WithMany().HasForeignKey(u => u.ContactId);
I personally would go with creating "mapper views" with CustomerId on legacy tables, as it's cleaner from c# layer perspective and you can make those views look like real tables. It is also difficult to suggest a solution without knowing what exactly is the scenario that you have a problem with: querying, loading, saving, etc.